| Re: [AD] About the Win2k Bug |
[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
> I would argue against merging it, from the sole basis of saying "if it
> ain't broke, don't fix it." I haven't looked at the code to see what kind
> of speed benefit you might get merging the threads.
Maybe only cutting the overhead due to thread switching.
> Even so, separate threads allow for better support for SMP and as was said
> earlier, do separate the code.
Yes, but a basic Allegro program uses no less than 5 threads ! Maybe 4
sounds more reasonable.
Regarding the code, we can maintain the modularity by implementing a
event/handler slot mechanism (my previously cited patch does that).
> If the threads have the capability to wait for their specific events, then
> merging them makes no sense whatsoever (if the mouse thread could use a
> WaitForMouseEvent() blocking call, then merging the threads makes little
> sense but much sense for separating that thread).
Quite right. The input threads are purely event-driven.
That's the main argument for not merging.
> Also, let me know when the problem is "supposedly fixed" in the cvs. I'll
> do a cvs update and make a package where I can test the program on a
> Win2k system but I'd rather not apply all of the individual patches on my
> system when I don't have 2k.
Henrik has already commited the patch. You can proceed.
--
Eric Botcazou
ebotcazou@xxxxxxxxxx